Python version of FLAME GPU 2

[daniele-baccega]

Hi,
what do you mean with transpiled in "agent functions expressed as Python syntax which are transpiled to C++"?

[Robadob]

Transpiling is the process of taking source code written in one language and transforming it into another language that has a similar level of abstraction.

In this case FLAMEGPU2's Python transpiler takes an agent function expressed in a pythonic syntax e.g.

@pyflamegpu.agent_function
def add_func(message_in: pyflamegpu.MessageNone, message_out: pyflamegpu.MessageNone):
    x = pyflamegpu.getVariableInt("x")
    pyflamegpu.setVariableInt("x", x + 20)
    return pyflamegpu.ALIVE

You pass it to the transpiler

        func_translated = pyflamegpu.codegen.translate(add_func)

This will store a string with the C++ equivalent agent function inside func_translated

e.g.

FLAMEGPU_AGENT_FUNCTION(add_func, flamegpu::MessageNone, flamegpu::MessageNone) {
    int x = FLAMEGPU->getVariable<int>("x");
    FLAMEGPU->setVariable<int>(x + 20);
    return flamegpu::ALIVE;
}

Which can then be passed to FLAMEGPU as a runtime compiled (RTC) function for use in a model. Or you could just look at the generated string, if you want to understand how the C++ equivalent would be structured.

func = a.newRTCFunction("add_func", func_translated)

There are some limitations to this, e.g. you can't start using numpy functions or similar python packages, although we support several numpy types as they map better to the full range of C/C++ data types available.

Hope that helps.

In drafting this reply, I spotted a mistake in one of the related code examples in the documentation, I will get that fixed ASAP.